AI-generated conceptual illustration. The portrait, person and writing are invented, not historical evidence; the arrangement is not an archival-storage demonstration.

A name pencilled on the back of a photograph can feel like a solved mystery. Turn it over and the face seems to acquire a place in the family tree.

Before you attach that portrait to a person, pause. The handwriting is evidence too. Who wrote it? When? Was the writer naming the sitter, the owner of the picture or the person who sent it?

A photograph can preserve a face remarkably well while leaving its identity uncertain. The Library of Congress guide to photographic evidence recommends evaluating photographs and captions as carefully as written records. [3]

Begin with the object you actually have

Give the photograph a simple identifier, such as PHOTO-014. Make a record of where it came from, who supplied it and whether you hold an original, a later print or a digital copy. Record the ownership history as far as you know it; gaps are useful information too. Photograph or scan the front, back and any mount or envelope, without cropping away edges or markings. Do not detach a photograph from a mount or album just to reveal its reverse. If the surface or binding is fragile, choose a method that protects the object and seek advice where needed. [4]

Transcribe the inscription exactly before interpreting it. Keep “Aunt Anna?” as “Aunt Anna?” rather than silently removing the question mark. If someone offers an identification, record their words, the date and the basis for their memory.

A caption added recently is different evidence from an inscription made when the photograph was sent. Both may help; neither should acquire certainty merely by being written down.

Separate seeing from concluding

Create two short columns in your notes: observation and interpretation.

“Three people standing beside a bicycle” belongs in the first. “A family celebrating a birthday” belongs in the second unless another source supports it. Similar clothing, close poses and apparent resemblance may suggest relationships, but they do not establish them on their own.

This habit is especially helpful when several relatives are reviewing the same picture. They can disagree about an identity while preserving the observations everyone can check.

Look past the faces

The Library of Congress offers a useful example of visual dating. In an undated office photograph, staff identified posters issued in 1918 and 1919. Those posters provided evidence that the scene was photographed in 1919 or later. The catalogue was updated to “1919?”—a tentative date, not a precisely established year. [1]

Apply that reasoning to your own material. An identifiable object may offer an earliest possible date. A studio address can become a research lead: look for contemporary directories or advertisements that show when the photographer used that address. A building in the background can send you to local records. Treat a business date range as something to test alongside the image, not an automatic date stamp; a mount or reproduction may have its own history. Clothing can suggest a range, but should not carry the whole conclusion.

Write down the limit of each clue. “No earlier than” is often a better result than an overconfident date.

Test a candidate identity outside the picture

If the inscription suggests an Anna, list the plausible Annas rather than choosing the first match. Compare their ages, locations, family connections and the history of the photograph’s ownership. Look for evidence that contradicts the proposed identification as well as evidence that supports it.

A second photograph with a reliable caption may be useful, but facial resemblance alone is a fragile bridge. If the case remains unresolved, keep the image as an unidentified photograph with a documented hypothesis. It still belongs in the family archive.

Build a small case instead of collecting agreeable clues

Imagine an inscription reading “Anna, Rotterdam”, with two possible Annas in the family. This is a fictional research exercise. One lived in Rotterdam as a young adult; the other visited relatives there decades later. The place-name alone cannot choose between them.

Make a candidate table in your research notes. For each Anna, record a birth date, documented places of residence, likely age in the image and the source for each fact. Then ask which evidence could distinguish them. A dated letter enclosing the photograph could be more useful than another relative saying that the face “looks right”.

Separate independent support from repeated information. Three online family trees carrying the same caption may all have copied one unverified upload. Count the underlying identification once, and trace it back to the person or document that first supplied it.

Look for contradictions before resolving them. If the proposed sitter died before the earliest supported date of the depicted scene, either the identity, the dating argument or your understanding of the image needs reconsideration. But a print made after someone died can still reproduce a photograph taken during their lifetime. Distinguish the scene or exposure date, the physical print date and the later scanning date. They answer different questions.

Ask relatives questions that preserve their memories

Show the whole photograph before offering your favourite name. Ask “What do you recognise?” and “How do you know?” rather than “That is Aunt Anna, isn’t it?” Record whether the person remembers the sitter directly, heard the name from someone else or is making a visual comparison now.

A confident answer deserves a source note, not automatic certainty. Preserve the contributor’s wording, relationship to the people shown and the date of the conversation. If two relatives disagree, keep both accounts with their reasons. That disagreement may point to a generation of names being mixed up, or to two similar photographs circulating in the family.

Give the next researcher something they can check

Use a short record alongside the image:

• Identifier and custody: PHOTO-014; who supplied it and where the original is kept.
• Exact inscription: including punctuation, uncertain letters and where the writing appears.
• Visible evidence: objects, signs, studio details and what is actually legible.
• Proposed identification: the candidate and the evidence supporting or contradicting it.
• Conclusion and next step: confirmed, probable, possible or unresolved, with an explanation of your wording and the next record to seek.

These labels are a suggested working vocabulary, not an official proof standard. Their value lies in the explanation behind them. “Probable” without a reason is merely a new caption.

Keep the unchanged source scan beside any edited version, with the method and software recorded. Ordinary sharpening or contrast adjustment can exaggerate edges, obscure tones or encourage a mistaken reading. AI reconstruction and some colourisation tools can go further by inventing plausible features or colours. Treat those results as interpretations, not recovered evidence. A clearer-looking badge, face or sign is not verified merely because software made it legible. Compare any proposed reading with the source and independent records. Before sharing a living person’s photograph publicly, discuss permission and sensitive information with the people concerned.

Preserve the original and the uncertainty

Make a clean, clear workspace and keep food and drink away. Handle photographs carefully at their edges, avoiding the image surface; the National Archives provides photograph-specific handling guidance. [5] This is distinct from its advice for ordinary paper documents.

For many prints, acid-free enclosures, suitable stable plastics and non-adhesive mounting can be useful. However, “archival” is not a guarantee that a product suits every photographic process or damaged surface. The National Archives’ family-archive storage guide and photograph-specific guidance are sensible starting points. [2, 5] A fragile, flaking, stuck or damaged original deserves conservation advice before removal or repair.

Keep the original after digitising it. A digital copy makes sharing easier, but it brings its own preservation risks. The National Archives recommends keeping three copies on two kinds of storage, with one copy off-site. [4] Check that your copies can actually be opened and restored; a filename in a folder is not proof of a usable backup.

Name related files consistently—for example, PHOTO-014-front, PHOTO-014-back and PHOTO-014-edited—and keep the identification notes with them. These names are suggested examples. Record the scan date as the scan date, not the date of the historical scene. A description can preserve both the current research conclusion and the uncertainty that remains.

A careful caption might read: “Possibly Anna de Vries; identification supplied by a relative in 2026, not yet corroborated.” This is an invented example of wording, not a claim about a real family. If research later changes the identification, date the revision and retain the reason for the earlier proposal.

Useful places to take the next research question

Match the resource to the question rather than searching every name everywhere.

• How should I assess a photograph? The Library of Congress photographic-evidence guide explains why captions and images need evaluation together. For a concrete example, read A Closer Look: Dating a Photo.
• Could this person have been in that place at that time? WieWasWie offers access to Dutch historical personal records. Use available records to test age, relationships and residence, then consult the underlying record and its archive reference. An index match alone does not identify a face. [6]
• Can I recognise a Dutch building or setting? The Rijksdienst voor het Cultureel Erfgoed image-bank guide leads to photographs and drawings of heritage, including buildings and landscapes. Compare structural details and dates; resemblance is a lead, not proof. [7]
• How can I make a copy without harming the original? Follow the National Archives’ digitising guidance, including its cautions about fragile material and forced bindings. [4]

When you use an online record, save the institution, collection, item identifier, description and link, together with the date you consulted it. Check each institution’s usage conditions before reproducing an image publicly. Your research notes should still make sense if a web address changes.

Your story is worth remembering. So is the trail of evidence that lets someone else understand how you reached it.

Sources and further reading

[1] Library of Congress — A Closer Look: Dating a Photo, Barbara Orbach Natanson, 7 September 2011. A worked example of an earliest possible scene date and a tentative catalogue date.

[2] US National Archives — Storing Family Archives. Enclosures, stable plastics and non-adhesive mounting.

[3] Library of Congress — Photographic Evidence. Evaluating photographs and captions as genealogical sources.

[4] US National Archives — Digitizing Family Papers and Photographs. Safe copying, retaining originals, metadata and backups.

[5] US National Archives — Photographs: Holdings Maintenance. Photograph-specific handling and housing considerations.

[6] WieWasWie — Dutch genealogical records. A research resource for testing biographical context, not facial identification.

[7] Rijksdienst voor het Cultureel Erfgoed — Beeldbank. A research resource for Dutch buildings and settings.